How to Prove Your Sales Training Worked

TL;DR

Leadership keeps asking whether the sales training budget moved win rates, and most enablement teams answer with completions and satisfaction scores, which is not the question.

The uncomfortable part: the stats the industry uses to justify itself are folklore. The famous “reps forget 87% of training within a month” line traces to a 2012 blog post with no study behind it.

The fix is not a better slide. It’s measuring behavior on real calls: define the behaviors that correlate with deals moving in your motion, baseline every rep, train, then watch whether the behaviors show up on live calls and whether the cohort’s deals move differently. That’s a system you can run, and it’s the only answer a CFO should accept.

I. The Stat Nobody Can Source

If you’ve sat through a sales training pitch in the last decade, you’ve seen the slide: “B2B reps forget 70% of training within a week and 87% within a month. (Gartner).”

We went looking for the study. There isn’t one. The earliest traceable version is a 2012 blog post from CEB’s Sales Executive Council, with no methodology, no sample, no data. Gartner bought CEB in 2017, the attribution upgraded itself to “Gartner,” and a blog post became a load-bearing fact for an industry. Its cousin, “84% of sales training is forgotten within 90 days,” is usually pinned to a report whose link has been dead since around 2016.

Sit with the irony for a second. The sales training industry cannot produce evidence for its own most-quoted claim about evidence. The people selling you measurement can’t cite their sources.

I’m not telling you this to dunk on vendors. I’m telling you because it’s the whole problem in miniature: this category runs on asserted outcomes instead of measured ones, and buyers have been trained to accept that. If your own board deck quotes that 87%, the discipline problem isn’t only on the vendor side.

II. What the Evidence Actually Says

Here’s what survives scrutiny.

Forgetting is real and it is fast. Hermann Ebbinghaus measured the curve in the 1880s, and Murre and Dros reproduced his results in 2015 in PLOS ONE: without reinforcement, most new information decays steeply within days, then the curve flattens. That’s memory science done on syllable lists, not a sales study, which is exactly why it’s credible. Nobody was selling anything.

And almost nobody measures whether training changed anything. The standard framework here is Kirkpatrick’s four levels, published in 1959 and still the field’s shared vocabulary: Level 1 is whether people liked the training, Level 2 whether they learned it, Level 3 whether their on-the-job behavior changed, Level 4 whether business results moved. Jack Phillips later added Level 5, converting results to ROI against the fully loaded cost. In ATD’s 2019 evaluation study of 779 talent-development organizations, 83% evaluated at Level 1 and 80% at Level 2, but only 54% evaluated any program at Level 3, and 16% ever reached ROI. And the organizations that did measure behavior applied it to about a third of their programs.

Put those two findings together and the picture is clear: decay is real, and the measurement that would catch it mostly doesn’t happen. Reaction scores at the exit survey, then silence. Which means most sales training programs are run the way you’d run a gym that weighs you only on the way in.

III. Why Your Current Dashboard Can’t Answer the Question

When a CRO asks “did the training work,” they are asking a Level 3 and Level 4 question: are reps doing something different on calls, and did deals move because of it. The enablement dashboard answers Level 1 and 2: completions, certifications, quiz scores, maybe a roleplay recording someone reviewed once.

The gap isn’t laziness. Until recently, Level 3 was unmeasurable at any real scale. Observing behavior change meant a manager sitting on calls, and a manager can sit on a handful of calls a month, chosen non-randomly, scored against whatever that manager happens to listen for. I ran sales at scale long enough to know exactly what that produces: confident narratives, thin evidence. (I wrote about the deeper version of this problem in why training can’t fix execution variance.)

So enablement teams did the rational thing: they reported what they could count. The cost shows up later, in the budget meeting where “4,200 modules completed” meets “win rate is flat” and loses.

IV. Measuring Behavior Change for Real

The mechanism that fixes this is not complicated, now that every call can be recorded and analyzed. It has four parts, and the order matters.

1. Define the behaviors from your outcomes, not from a curriculum. Don’t start with what the training taught. Start with what your won and lost deals say actually matters: the discovery question that correlates with second meetings, the pricing framing that survives procurement, the follow-up speed on champion emails. When we run this analysis for teams, the list usually surprises them. At one customer the single behavior most correlated with deals progressing was assumptive closing, which nobody had ever coached because it wasn’t on anyone’s scorecard. A scorecard checks what someone already believed mattered. Outcomes tell you what does.

2. Baseline every rep before the training. Behavior change is a delta, and you can’t report a delta you never baselined. This is per rep, not team average: your top rep already does the behavior, your newest rep doesn’t, and the training’s job is different for each. Sample sizes matter here, which is why observing every call beats a manager’s monthly ride-along by an order of magnitude.

3. Watch the behavior on live calls, not in the certification roleplay. The certification tells you a rep can perform the behavior with their manager watching. Level 3 is whether they do perform it three weeks later with a real buyer pushing back. That’s the measurement that makes the forgetting curve visible: if adoption spikes in week one and decays to baseline by week four, you’ve learned your program delivers knowledge but not habit, and reinforcement, practice against the objections your reps actually lose to, is where the next dollar should go. Measurement without reinforcement just documents the decay.

4. Tie the behavior to outcomes, cohort against cohort. Now the Level 4 question is answerable in a form a CFO respects: deals where the trained behaviors showed up versus deals where they didn’t, trained cohort versus not-yet-trained cohort, stage progression and win rate over the following quarter. Be honest about attribution. You’re not running a clinical trial; markets shift and territories differ. But “the reps who adopted the behavior won more, the reps who didn’t, didn’t” is evidence. “Everyone completed the course” is not.

The pattern behind all four steps: the unit of measurement is one rep’s behavior on one real conversation. Everything your current stack reports is an aggregate built to hide exactly that.

V. What to Put in Front of Leadership

The QBR version of this fits on one slide, and none of it is a satisfaction score.

• Behavior adoption, per behavior, per rep cohort, on live calls: your leading indicator, visible within weeks.
• Retention of adoption at 30, 60, 90 days: the honest replacement for the fake forgetting stat, measured on your own team.
• Outcome deltas for adopters versus non-adopters: stage progression, win rate, ramp time for new hires.
• Coaching coverage: what share of reps got specific, call-based feedback this month. In most orgs the honest number is “the two reps the manager had time for.” CSO Insights’ enablement research spent years associating formal coaching discipline with meaningfully higher win rates; the direction of that finding matches everything we see in call data, and coverage is the input you control.

Notice what this reframes. Training stops being an event you bought and becomes a system you operate: define, baseline, train, observe, reinforce, attribute. The vendors who live on the folklore slide hate this framing, because it makes their outcomes checkable. That is precisely why you should adopt it, for them and for us. If a tool, ours included, claims it changes what reps do on calls, the evidence should be sitting in your own call data within a quarter.

The money at stake isn’t small. US companies spend on the order of $100 billion a year on corporate training (Training magazine’s 2025 industry report puts it at $102.8 billion), and the sales slice of that is bought almost entirely on Level 1 evidence. You don’t need a better stat to justify the spend. You need to stop needing stats that someone else can’t source, because the only decay curve that matters is the one measurable on your own team’s calls.

Start there. Pick the three behaviors your best reps do and your pipeline proves out, baseline them this month, and let the next training investment walk into the QBR with its own receipts.